Comparing exploratory and confirmatory factor analysis: A study on the 5-factor model of personality

Summary-Several authors claim that widespread support exists for a 5-factor model of personality ratings. In the present study, structural equation modeling was used to investigate this issue. The subjects (128 males and 128 females) were administered Costa and McCrae’s NE0 Personality Inventory, and they rated themselves, and were rated by three acquaintances, on the 20 adjective scales suggested by Norman as marker variables for the Big Five. Coefficients of factor comparability indicated that a Sfactor model accounted for the data better than any other model. Moreover, the five factors that were obtained matched conventional measures of the Big Five very well. A multitrait-multimethod analysis with five traits and three methods yielded acceptable convergent and discriminant validities, and a model with oblique trait as well as oblique method factors was supported by structural equation models. A confirmatory factor-analytic model, however, that predicted the correlations among 60 variables from five trait factors and three method factors, did not fit the data. It is concluded that this finding reflects a desirable heterogeneity of personality factors as higher-level constructs. The implications for the usefulness of confirmatory factor analysis as well as for the Sfactor model of personality are discussed